worklog-manager

To bootstrap Codex worklog context, initialize the .agents/worklog/codex/{plan,todo,learn} directories. This skill automates that setup and validates the learn_index.md startup state to ensure a consistent baseline for new sessions.

Learn metadata fails consistency checks when frontmatter is missing required fields like status, freshness, or last_validated_at. This skill audits learn entries and enforces these requirements to detect drift-prone entries before you trust prior sessions.

The best way to audit learn_index.md coherence is running the provided audit script. It validates per-learn metadata, checks frontmatter requirements including review_after when applicable, and ensures index/file coherence across .agents/worklog/codex.

Yes, you can detect drift-prone entries by running a startup audit. The skill generates deterministic startup summaries for active learnings and flags entries that lack required frontmatter or fail consistency validation against the learn index.

No dependencies are required to validate learn metadata frontmatter. The skill operates independently using bundled scripts and references to enforce required fields like status, freshness, and last_validated_at across your worklog entries.

When bootstrapping worklog context, the main limitation is trusting prior sessions without auditing. The skill requires valid frontmatter and index/file coherence; entries missing review_after or last_validated_at are flagged as drift-prone and cannot be trusted automatically.